Hi Team, With several team members we are working on optimizing ACID transaction related metastore calls. Specifically aiming to have non-blocking openTxns calls (so 2 parallel openTxns can run without blocking each other) which could increase the throughput of Hive a lot.
Currently based on this document (https://cwiki.apache.org/confluence/display/Hive/AdminManual+Metastore+Administration <https://cwiki.apache.org/confluence/display/Hive/AdminManual+Metastore+Administration>) we support: MySQL 5.6.17 mysql Postgres 9.1.13 postgres Oracle 11g oracle hive.metastore.orm.retrieveMapNullsAsEmptyStrings <https://cwiki.apache.org/confluence/display/Hive/Configuration+Properties#ConfigurationProperties-hive.metastore.orm.retrieveMapNullsAsEmptyStrings> MS SQL Server 2008 R2 mssql All of the databases above support one way or another to generate IDENTITY/AUTO INCREMENT values, but Oracle 11g. We could use different SQL queries for Oracle 11g (like SEQUENCE + NEXT_VAL), but that would mean that we have to generate different queries for Oracle backward compatibility. Oracle 11g Extended support due to expire in 31st December 2020, see: https://www.oracle.com/webfolder/community/oracle_database/3905940.html <https://www.oracle.com/webfolder/community/oracle_database/3905940.html> I do not see too much overlap between Oracle 11g, and Hive 4.0.0. Since Oracle 12c supports IDENTITY columns as well, I propose that for Hive 4.0.0 we support only Oracle 12c instead of adding quickly outdated complexity. Any thoughts, ideas are welcome. Thanks, Peter
